Premium
Relating occlusion maps obtained through deep learning to functional impairment in dementia of Alzheimer’s type
Author(s) -
Bae Jinhyeong,
Stocks Jane,
Heywood Ashley,
Jung Youngmoon,
Katsaggelos Aggelos,
Jenkins Lisanne M,
Karteek Popuri,
Beg Mirza Faisal,
Wang Lei
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.043538
Subject(s) - voxel , dementia , psychology , artificial intelligence , clinical dementia rating , occlusion , correlation , cognition , audiology , nuclear medicine , cognitive impairment , medicine , computer science , neuroscience , mathematics , disease , geometry
Background Predicting the conversion from Mild Cognitive Impairment (MCI) into Dementia of the Alzheimer’s type (DAT) and functional change is crucial to patient care and treatment. In order to visualize brain regions which are significant in the prediction, we implemented an occlusion map based on deep learning. Method Using T1‐weighted structural MRI data from ADNI, 3D convolutional neural network was trained to predict the conversion from MCI to DAT through a transfer learning pipeline. The model resulted in an 82.4% classification accuracy on an independent test set. An occlusion map was subsequently generated as follows. Each brain scan was occluded by 2x2x2 voxel patch iterated through every position in the brain. The model produced a prediction score corresponding to the location of the occlusion patch to identify important voxels for the model’s prediction at the subject level. Mean intensity value within the occlusion map was obtained in Gray Matter. This was used to correlate to clinical scores’ rate of change, which include Clinical Dementia Rating‐Sum of Boxes (CDRSB), Alzheimer’s Disease Assessment Scale – cognitive 11 item (ADAS11) and cognitive 13 item (ADAS13), Mini Mental State Exam (MMSE), Rey Auditory Verbal Learning Test (RAVLT) – RAVLT Immediate (IMD), RAVLT Learning (LRN), RAVLT Forgetting (FRG), RAVLT Percent Forgetting (PCF), and Functional Activities Questionnaire (FAQ). Result The occlusion map identified regions important to the prediction of conversion. Mean intensity values of the gray matter within the occlusion map showed significant correlation with behavior measures: decreased intensity (indicating gray matter loss) was associated with decreased memory performance as measured by RAVLT immediate recall performance, increased cognitive dysfunction as measured by ADAS11, ADAS13, and increased daily living deficits as measured by FAQ. Conclusion These results indicate brain regions associated with cognitive change during conversion from MCI to DAT. These regions provide validity of the deep learning model’s results and provide insights on patient functional changes during conversion.